Machine learning models and performance dependency on 2D chemical descriptor space for retention time prediction of pharmaceuticals.

The predictive modeling of liquid chromatography methods can be an invaluable asset, potentially saving countless hours of labor while also reducing solvent consumption and waste. Tasks such as physicochemical screening and preliminary method screening systems where large amounts of chromatography data are collected from fast and routine operations are particularly well suited for both leveraging large datasets and benefiting from predictive models. Therefore, the generation of predictive models for retention time is an active area of development. However, for these predictive models to gain acceptance, researchers first must have confidence in model performance and the computational cost of building them should be minimal. In this study, a simple and cost-effective workflow for the development of machine learning models to predict retention time using only Molecular Operating Environment 2D descriptors as input for support vector regression is developed. Furthermore, we investigated the relative performance of models based on molecular descriptor space by utilizing uniform manifold approximation and projection and clustering with Gaussian mixture models to identify chemically distinct clusters. Results outlined herein demonstrate that local models trained on clusters in chemical space perform equivalently when compared to models trained on all data. Through 10-fold cross-validation on a comprehensive set containing 67,950 of our company's proprietary analytes, these models achieved coefficients of determination of 0.84 and 3 % error in terms of retention time. This promising statistical significance is found to translate from cross-validation to prospective prediction on an external test set of pharmaceutically relevant analytes. The observed equivalency of global and local modeling of large datasets is retained with METLIN's SMRT dataset, thereby confirming the wider applicability of the developed machine learning workflows for global models.

Bioisostere Effects on the EPSA of Common Permeability-Limiting Groups.

Polar molecular surface area provides a valuable metric when optimizing properties as varied as membrane permeability and efflux susceptibility. The EPSA method to measure this quantity has had a substantial impact in medicinal chemistry, providing insight into the conformational and stereoelectronic features that govern the polarity of small molecules, targeted protein degraders, and macrocyclic peptides. Recognizing the value of bioisosteres in replacing permeation-limiting polar groups, we determined the effects of common amide, carboxylic acid, and phenol bioisosteres on EPSA, using matched molecular pairs within the Merck compound collection. Our findings reinforce EPSA's utility in optimizing permeability, highlight bioisosteres within each class that are particularly effective in lowering EPSA and others, which despite widespread use, offer little to no such benefit. Our method for matched-pair identification is generalizable across large compound collections and, thus, may constitute a flexible platform to study the effects of bioisosterism both in EPSA and other in vitro assays.

Discovery of MK-8153, a Potent and Selective ROMK Inhibitor and Novel Diuretic/Natriuretic.

A renal outer medullary potassium channel (ROMK, Kir1.1) is a putative drug target for a novel class of diuretics with potential for treating hypertension and heart failure. Our first disclosed clinical ROMK compound, 2 (MK-7145), demonstrated robust diuresis, natriuresis, and blood pressure lowering in preclinical models, with reduced urinary potassium excretion compared to the standard of care diuretics. However, 2 projected to a short human half-life (∼5 h) that could necessitate more frequent than once a day dosing. In addition, a short half-life would confer a high peak-to-trough ratio which could evoke an excessive peak diuretic effect, a common liability associated with loop diuretics such as furosemide. This report describes the discovery of a new ROMK inhibitor 22e (MK-8153), with a longer projected human half-life (∼14 h), which should lead to a reduced peak-to-trough ratio, potentially extrapolating to more extended and better tolerated diuretic effects.

Octanol-water partition coefficient measurements for the SAMPL6 blind prediction challenge.

Partition coefficients describe the equilibrium partitioning of a single, defined charge state of a solute between two liquid phases in contact, typically a neutral solute. Octanol-water partition coefficients ([Formula: see text]), or their logarithms (log P), are frequently used as a measure of lipophilicity in drug discovery. The partition coefficient is a physicochemical property that captures the thermodynamics of relative solvation between aqueous and nonpolar phases, and therefore provides an excellent test for physics-based computational models that predict properties of pharmaceutical relevance such as protein-ligand binding affinities or hydration/solvation free energies. The SAMPL6 Part II octanol-water partition coefficient prediction challenge used a subset of kinase inhibitor fragment-like compounds from the SAMPL6 [Formula: see text] prediction challenge in a blind experimental benchmark. Following experimental data collection, the partition coefficient dataset was kept blinded until all predictions were collected from participating computational chemistry groups. A total of 91 submissions were received from 27 participating research groups. This paper presents the octanol-water log P dataset for this SAMPL6 Part II partition coefficient challenge, which consisted of 11 compounds (six 4-aminoquinazolines, two benzimidazole, one pyrazolo[3,4-d]pyrimidine, one pyridine, one 2-oxoquinoline substructure containing compounds) with log P values in the range of 1.95-4.09. We describe the potentiometric log P measurement protocol used to collect this dataset using a Sirius T3, discuss the limitations of this experimental approach, and share suggestions for future log P data collection efforts for the evaluation of computational methods.

pKa measurements for the SAMPL6 prediction challenge for a set of kinase inhibitor-like fragments.

Determining the net charge and protonation states populated by a small molecule in an environment of interest or the cost of altering those protonation states upon transfer to another environment is a prerequisite for predicting its physicochemical and pharmaceutical properties. The environment of interest can be aqueous, an organic solvent, a protein binding site, or a lipid bilayer. Predicting the protonation state of a small molecule is essential to predicting its interactions with biological macromolecules using computational models. Incorrectly modeling the dominant protonation state, shifts in dominant protonation state, or the population of significant mixtures of protonation states can lead to large modeling errors that degrade the accuracy of physical modeling. Low accuracy hinders the use of physical modeling approaches for molecular design. For small molecules, the acid dissociation constant (pKa) is the primary quantity needed to determine the ionic states populated by a molecule in an aqueous solution at a given pH. As a part of SAMPL6 community challenge, we organized a blind pKa prediction component to assess the accuracy with which contemporary pKa prediction methods can predict this quantity, with the ultimate aim of assessing the expected impact on modeling errors this would induce. While a multitude of approaches for predicting pKa values currently exist, predicting the pKas of drug-like molecules can be difficult due to challenging properties such as multiple titratable sites, heterocycles, and tautomerization. For this challenge, we focused on set of 24 small molecules selected to resemble selective kinase inhibitors-an important class of therapeutics replete with titratable moieties. Using a Sirius T3 instrument that performs automated acid-base titrations, we used UV absorbance-based pKa measurements to construct a high-quality experimental reference dataset of macroscopic pKas for the evaluation of computational pKa prediction methodologies that was utilized in the SAMPL6 pKa challenge. For several compounds in which the microscopic protonation states associated with macroscopic pKas were ambiguous, we performed follow-up NMR experiments to disambiguate the microstates involved in the transition. This dataset provides a useful standard benchmark dataset for the evaluation of pKa prediction methodologies on kinase inhibitor-like compounds.

Comparing Dog and Human Intestinal Fluids: Implications on Solubility and Biopharmaceutical Risk Assessment.

Despite many documented differences in gut physiology compared to humans, the beagle dog has been successfully used as a preclinical model for assessing the relative bioavailability of dosage forms during formulation development. However, differences in pH and bile salt concentration and micellar structure between dog and human intestinal fluids may influence the solubility and dissolution behavior of especially BCS II/IV compounds. Recently, a canine fasted simulated intestinal fluid (FaSSIFc) mimicking the composition in the lumen of the beagle dog under the fasted state has been proposed. In this manuscript, we present the utilization of FaSSIFc to compare solubility of several preclinical candidates against human FaSSIF. While solubility of free bases and neutral compounds was easily predicted by the relative amounts of sodium taurocholate in the fluids, free acids were shown to be much more soluble in FaSSIFc owing to both the solubility at higher pH as well as the increased bile salt concentration. For one of the model compounds, we demonstrate that the high solubility necessitates the need for a formulation comparison at a relatively higher dose in the dog to mimic the outcome of a human relative bioavailability study. Finally, we show how using the solubility value in FaSSIFc for the same compound results in better predictability of the plasma concentration profiles in dogs from a physiologically based absorption model. The collective data indicate that caution and more detailed measurements are required if the dog is used as the preclinical model for the development of formulations of weak acids.

Discovery of Vibegron: A Potent and Selective ß3 Adrenergic Receptor Agonist for the Treatment of Overactive Bladder.

The discovery of vibegron, a potent and selective human ß3-AR agonist for the treatment of overactive bladder (OAB), is described. An early-generation clinical ß3-AR agonist MK-0634 (3) exhibited efficacy in humans for the treatment of OAB, but development was discontinued due to unacceptable structure-based toxicity in preclinical species. Optimization of a series of second-generation pyrrolidine-derived ß3-AR agonists included reducing the risk for phospholipidosis, the risk of formation of disproportionate human metabolites, and the risk of formation of high levels of circulating metabolites in preclinical species. These efforts resulted in the discovery of vibegron, which possesses improved druglike properties and an overall superior preclinical profile compared to MK-0634. Structure-activity relationships leading to the discovery of vibegron and a summary of its preclinical profile are described.

Use of molecular modeling aided design to dial out hERG liability in adenosine A(2A) receptor antagonists.

Molecular modeling was performed on a triazolo quinazoline lead compound to help develop a series of adenosine A2A receptor antagonists with improved hERG profile. Superposition of the lead compound onto MK-499, a benchmark hERG inhibitor, combined with pKa calculations and measurement, identified terminal fluorobenzene to be responsible for hERG activity. Docking of the lead compound into an A2A crystal structure suggested that this group is located at a flexible, spacious, and solvent-exposed opening of the binding pocket, making it possible to tolerate various functional groups. Transformation analysis (MMP, matched molecular pair) of in-house available experimental data on hERG provided suggestions for modifications in order to mitigate this liability. This led to the synthesis of a series of compounds with significantly reduced hERG activity. The strategy used in the modeling work can be applied to other medicinal chemistry programs to help improve hERG profile.

Interaction of cholesterol-conjugated ionizable amino lipids with biomembranes: lipid polymorphism, structure-activity relationship, and implications for siRNA delivery.

Delivery of siRNA is a major obstacle to the advancement of RNAi as a novel therapeutic modality. Lipid nanoparticles (LNP) consisting of ionizable amino lipids are being developed as an important delivery platform for siRNAs, and significant efforts are being made to understand the structure-activity relationship (SAR) of the lipids. This article uses a combination of small-angle X-ray scattering (SAXS) and differential scanning calorimetry (DSC) to evaluate the interaction between cholesterol-conjugated ionizable amino lipids and biomembranes, focusing on an important area of lipid SAR--the ability of lipids to destabilize membrane bilayer structures and facilitate endosomal escape. In this study, cholesterol-conjugated amino lipids were found to be effective in increasing the order of biomembranes and also highly effective in inducing phase changes in biological membranes in vitro (i.e., the lamellar to inverted hexagonal phase transition). The phase transition temperatures, determined using SAXS and DSC, serve as an indicator for ranking the potency of lipids to destabilize endosomal membranes. It was found that the bilayer disruption ability of amino lipids depends strongly on the amino lipid concentration in membranes. Amino lipids with systematic variations in headgroups, the extent of ionization, tail length, the degree of unsaturation, and tail asymmetry were evaluated for their bilayer disruption ability to establish SAR. Overall, it was found that the impact of these lipid structure changes on their bilayer disruption ability agrees well with the results from a conceptual molecular "shape" analysis. Implications of the findings from this study for siRNA delivery are discussed. The methods reported here can be used to support the SAR screening of cationic lipids for siRNA delivery, and the information revealed through the study of the interaction between cationic lipids and biomembranes will contribute significantly to the design of more efficient siRNA delivery vehicles.

Ionization behavior of amino lipids for siRNA delivery: determination of ionization constants, SAR, and the impact of lipid pKa on cationic lipid-biomembrane interactions.

Ionizable amino lipids are being pursued as an important class of materials for delivering small interfering RNA (siRNA) therapeutics, and research is being conducted to elucidate the structure-activity relationships (SAR) of these lipids. The pK(a) of cationic lipid headgroups is one of the critical physiochemical properties of interest due to the strong impact of lipid ionization on the assembly and performance of these lipids. This research focused on developing approaches that permit the rapid determination of the relevant pK(a) of the ionizable amino lipids. Two distinct approaches were investigated: (1) potentiometric titration of amino lipids dissolved in neutral surfactant micelles; and (2) pH-dependent partitioning of a fluorescent dye to cationic liposomes formulated from amino lipids. Using the approaches developed here, the pK(a) values of cationic lipids with distinct headgroups were measured and found to be significantly lower than calculated values. It was also found that lipid-lipid interaction has a strong impact on the pK(a) values of lipids. Lysis of model biomembranes by cationic lipids was used to evaluate the impact of lipid pK(a) on the interaction between cationic lipids and cell membranes. It was found that cationic lipid-biomembrane interaction depends strongly on lipid pK(a) and solution pH, and this interaction is much stronger when amino lipids are highly charged. The presence of an optimal pK(a) range of ionizable amino lipids for siRNA delivery was suggested based on these results. The pK(a) methods reported here can be used to support the SAR screen of cationic lipids for siRNA delivery, and the information revealed through studying the impact of pK(a) on the interaction between cationic lipids and cell membranes will contribute significantly to the design of more efficient siRNA delivery vehicles.

Difluoroethylamines as an amide isostere in inhibitors of cathepsin K.

The trifluoroethylamine group found in cathepsin K inhibitors like odanacatib can be replaced by a difluoroethylamine group. This change increased the basicity of the nitrogen which positively impacted the log D. This translated into an improved oral bioavailability in pre-clinical species. Difluoroethylamine compounds exhibit a similar potency against cathepsin K and selectivity profile against other cathepsins when compared to trifluoroethylamine analogs.

Design and synthesis of a new class of malonyl-CoA decarboxylase inhibitors with anti-obesity and anti-diabetic activities.

A new series of thiazole-substituted 1,1,1,3,3,3-hexafluoro-2-propanols were prepared and evaluated as malonyl-CoA decarboxylase (MCD) inhibitors. Key analogs caused dose-dependent decreases in food intake and body weight in obese mice. Acute treatment with these compounds also led to a drop in elevated blood glucose in a murine model of type II diabetes.

Discovery and X-ray crystallographic analysis of a spiropiperidine iminohydantoin inhibitor of beta-secretase.

A high-throughput screen at 100 microM inhibitor concentration for the BACE-1 enzyme revealed a novel spiropiperidine iminohydantoin aspartyl protease inhibitor template. An X-ray cocrystal structure with BACE-1 revealed a novel mode of binding whereby the inhibitor interacts with the catalytic aspartates via bridging water molecules. Using the crystal structure as a guide, potent compounds with good brain penetration were designed.

Discovery of N-{(1S,2S)-2-(3-cyanophenyl)- 3-[4-(2-[18F]fluoroethoxy)phenyl]-1-methylpropyl}- 2-methyl-2-[(5-methylpyridin-2-yl)oxy]propanamide, a cannabinoid-1 receptor positron emission tomography tracer suitable for clinical use.

The discovery of a structurally distinct cannabinoid-1 receptor (CB1R) positron emission tomography tracer is described. Starting from an acyclic amide CB1R inverse agonist (1) as the lead compound, an efficient route to introduce 18F to the molecule was developed. Further optimization focused on reducing the lipophilicity and increasing the CB1R affinity. These efforts led to the identification of [18F]-16 that exhibited good brain uptake and an excellent signal-to-noise ratio in rhesus monkeys.

Cyclopentane-based human NK1 antagonists. Part 1: discovery and initial SAR.

An initial investigation of the novel cyclopentane scaffold 6 afforded low nanomolar human NK1 antagonists having enhanced water solubility properties compared to morpholine 1. A synthesis of this cyclopentane scaffold, having three contiguous chiral centers, and the unexpected determination that the 1,2-trans-2,3-trans-ring stereochemistry, as opposed to the cis-ether/phenyl configuration of the known structures 1-5, is optimal for this class of antagonist are described.

(2S,3S)-3-Amino-4-(3,3-difluoropyrrolidin-1-yl)-N,N-dimethyl-4-oxo-2-(4-[1,2,4]triazolo[1,5-a]-pyridin-6-ylphenyl)butanamide: a selective alpha-amino amide dipeptidyl peptidase IV inhibitor for the treatment of type 2 diabetes.

A series of beta-substituted biarylphenylalanine amides were synthesized and evaluated as inhibitors of dipeptidyl peptidase IV (DPP-4) for the treatment of type 2 diabetes. Optimization of the metabolic profile of early analogues led to the discovery of (2S,3S)-3-amino-4-(3,3-difluoropyrrolidin-1-yl)-N,N-dimethyl-4-oxo-2-(4-[1,2,4]triazolo[1,5-a]pyridin-6-ylphenyl)butanamide (6), a potent, orally active DPP-4 inhibitor (IC(50) = 6.3 nM) with excellent selectivity, oral bioavailability in preclinical species, and in vivo efficacy in animal models. Compound 6 was selected for further characterization as a potential new treatment for type 2 diabetes.

Substituted 3-amino biaryl propionic acids as potent VLA-4 antagonists.

A series of substituted N-(3,5-dichlorobenzenesulfonyl)-(L)-prolyl- and (L)-azetidyl-beta-biaryl beta-alanine derivatives was prepared as selective and potent VLA-4 antagonists. The 2,6-dioxygenated biaryl substitution pattern is important for optimizing potency. Oral bioavailability was variable and may be a result of binding to circulating plasma proteins.

Substituted N-(3,5-dichlorobenzenesulfonyl)-L-prolyl-phenylalanine analogues as potent VLA-4 antagonists.

A series of substituted N-(3,5-dichlorobenzenesulfonyl)-L-prolyl- and alpha-methyl-L-prolyl-phenylalanine derivatives was prepared as VLA-4/VCAM antagonists. The compounds showed excellent potency with a wide variety of neutral, polar, electron withdrawing or donating groups on the phenylalanine ring (IC50 approximately 1 nM). Heteroaryl ring substitution for phenylalanine was also well tolerated. Pharmacokinetic studies in rat were performed on a representative set of compounds in both series.